A Deep Learning Functional Estimator of Optimal Dynamics for Sampling Large Deviations

Abstract

In stochastic systems, numerically sampling the relevant trajectories for the estimation of the large deviation statistics of time-extensive observables requires overcoming their exponential (in space and time) scarcity. The optimal way to access these rare events is by means of an auxiliary dynamics obtained from the original one through the so-called ``generalised Doob transformation''. While this optimal dynamics is guaranteed to exist its use is often impractical, as to define it requires the often impossible task of diagonalising a (tilted) dynamical generator. While approximate schemes have been devised to overcome this issue they are difficult to automate as they tend to require knowledge of the systems under study. Here we address this problem from the perspective of deep learning. We devise an iterative semi-supervised learning scheme which converges to the optimal or Doob dynamics with the clear advantage of requiring no prior knowledge of the system. We test our method in a paradigmatic statistical mechanics model with non-trivial dynamical fluctuations, the fully packed classical dimer model on the square lattice, showing that it compares favourably with more traditional approaches. We discuss broader implications of our results for the study of rare dynamical trajectories.

arXiv comment12 pages, 5 figures

References in corpus (40)

• The large deviation approach to statistical mechanics
• A high-bias, low-variance introduction to Machine Learning for physicists
• Dynamic first-order phase transition in kinetically constrained models of glasses
• The "Coulomb phase" in frustrated systems
• First-order dynamical phase transition in models of glasses: an approach based on ensembles of histories
• Fluctuations and response of nonequilibrium states
• Thermodynamic formalism for systems with Markov dynamics
• Direct evaluation of large-deviation functions
• Nonequilibrium Markov processes conditioned on large deviations
• Accelerate Monte Carlo Simulations with Restricted Boltzmann Machines
• Large deviations and ensembles of trajectories in stochastic models
• Space-time Thermodynamics of the Glass Transition
• Nonequilibrium microcanonical and canonical ensembles and their equivalence
• Simulating rare events in dynamical processes
• Canonical structure of dynamical fluctuations in mesoscopic nonequilibrium steady states
• Detection of phase transition via convolutional neural network
• Aspects of non-equilibrium in classical and quantum systems: slow relaxation and glasses, dynamical large deviations, quantum non-ergodicity, and open quantum dynamics
• A numerical approach to large deviations in continuous-time
• Hyperuniformity and phase separation in biased ensembles of trajectories for diffusive systems
• Frenesy: time-symmetric dynamical activity in nonequilibria
• Population dynamics method with a multi-canonical feedback control
• Making rare events typical in Markovian open quantum systems
• Dynamical phase transition for current statistics in a simple driven diffusive system
• Self-learning Monte Carlo with Deep Neural Networks
• Using matrix product states to study the dynamical large deviations of kinetically constrained models
• DMRG-study of current and activity fluctuations near non-equilibrium phase transitions
• Exact fluctuations of nonequilibrium steady states from approximate auxiliary dynamics
• Super-resolving the Ising model with convolutional neural networks
• Self-Learning Monte Carlo Method: Continuous-Time Algorithm
• Rare behavior of growth processes via umbrella sampling of trajectories
• Dynamic transition in an atomic glass former: a molecular dynamics evidence
• Importance sampling large deviations in nonequilibrium steady states. I
• Adaptive sampling of large deviations
• Fluctuating observation time ensembles in the thermodynamics of trajectories
• Dynamical phase behavior of the single- and multi-lane asymmetric simple exclusion process via matrix product states
• Activity phase transition for constrained dynamics
• Recommender Engine for Continuous Time Quantum Monte Carlo Methods
• Phase transition encoded in neural network
• Phases of quantum dimers from ensembles of classical stochastic trajectories
• Direct evaluation of dynamical large-deviation rate functions using a variational ansatz

Cited by in corpus (15)

• Machine Learning for Quantum Matter
• Phase Diagram of Active Brownian Spheres: Crystallization and the Metastability of Motility-Induced Phase Separation
• A reinforcement learning approach to rare trajectory sampling
• Learning nonequilibrium control forces to characterize dynamical phase transitions
• Reinforcement learning of rare diffusive dynamics
• Finite time large deviations via matrix product states
• Optimal sampling of dynamical large deviations via matrix product states
• Evolutionary reinforcement learning of dynamical large deviations
• Dynamical large deviations of two-dimensional kinetically constrained models using a neural-network state ansatz
• Phase Coexistence Implications of Violating Newton's Third Law
• Generating discrete-time constrained random walks and Lévy flights
• Modeling of biomolecular machines in non-equilibrium steady states
• Finding the effective dynamics to make rare events typical in chaotic maps
• Adaptive power method for estimating large deviations in Markov chains
• A unified perspective on exponential tilt and bridge algorithms for rare trajectories of discrete Markov processes